Azure Machine Learning Service is a fully managed cloud service that is used to train, deploy, and manage machine learning models.
In this post, we are going to see what an Azure Machine Learning service is and describe the end-to-end workflow for a machine learning service.
This post is beneficial for those who are new to machine learning or for all those who are preparing for Microsoft Azure Data Scientist Certification [DP-100].
Azure Machine Learning Service Overview
Azure Machine Learning Service is an enterprise-level service for building and deploying machine learning models. It allows us to create, test, manage, deploy, or monitor ML models in a scalable cloud-based environment. It supports numerous open-source packages available in Python such as TensorFlow, Matplotlib, and scikit-learn.
There are some tools available for Azure Machine Learning Service like which can be used to build, train, and deploy models.
- Azure Machine Learning Studio: It is a workspace where you create, build, train the machine learning models. To know more about Azure Machine Learning studio you can read our blog post on Azure Machine Learning Studio.
- Azure Machine Learning for Visual Studio Code Extension: It is a free extension that allows managing resources, model training workflows, and deployments in Visual Studio Code
- Jupyter Notebooks: It is an open-source web application that allows us to create and share documents that contain live code, visualizations narrative texts, and equations.
- Model Registry: It is a component of Azure Machine Learning Service where the model is stored once trained. A model registry is responsible for keeping records of the models that are being built and trained. The models can be identified by their versions and names. Every time a new model is registered with the name which has already been used before, the registry stores it as a new version. The version number is increased and the name of the model remains the same. Additional metadata tags can be added during the registration of the model which helps in easy searching.
- Image Registry: It holds a record of images created by the models. It adds additional metadata tags while creating an image that is kept by an image registry. These tags can be used as a query to find the image.
Features Of Azure Machine Learning Service
- Azure Machine Learning Service has the potential to auto-train and auto-tunes a model.
- The model can be trained on a local machine and then deployed on the cloud.
- It offers computing services like Azure Databricks, Azure Machine Learning Compute, etc.
- It manages the scripts and the run history of models, making it easy to compare model versions.
Workflow Of Azure Machine Learning Service
Azure machine learning service workflow is a three-step process that includes:
- Prepare Data
- Experiment (Build, Train & Test the model)
- Deployment
Before we start with collecting and processing our data we need a Workspace where we can perform all the operations. A Workspace represents the highest level centralized resource of Azure Machine Learning service. It holds the list of all compute targets used for the training developed model. It stores the log of training execution, metrics, outputs, and snapshots. This data assists in choosing the best training model for the project. The model is registered through the workspace.
1.) Prepare Data
This is the first step in creating a machine learning model which includes collecting and processing the data from datastore and datasets.
Datastore: They are used to store connection information to Azure storage services which can be referred by name and are attached to the workspace.
Some examples of supported Azure storage services that can be registered as datastores are:
- Azure Data Lake
- Azure SQL Database
- Databricks File System
- Azure Blob Container
Datasets: A Dataset is a reference to data in the datastore or behind public web URLs and also creates a copy of its metadata. There are two types of datasets supported by Azure namely the File dataset and Tabular dataset.
2.) Experiment
After the data is registered and stored in the dataset, the next step is to build, train, and test the model.
Model: It is a piece of code that takes input and produces the output for the given inputs. While developing a machine learning model, it requires selecting an algorithm, availing data, and tuning of hyperparameters. Training includes an iterative process which provides a trained model inheriting what it learned from the training process. The model is obtained by executing in Azure Machine Learning. To know more about Machine Learning models you can read our blog post on Azure ML Model.
Compute Targets: It is a machine or a set of machines that are used to run the training scripts or host service deployments. A local machine or a remote compute resource can also be used as a compute target. The compute resources used for compute targets are attached to the workspace.
There are 4 types of Compute Targets:
- Local Compute: It is a compute context where experiment submission code runs.
- Compute Cluster: It is a virtual cluster managed by Azure Machine Learning.
- Inference Cluster: It is a container-based deployment target.
- Attached Compute: It includes Azure Databricks, Azure Data Analytics, etc.
3.) Deployment
Once the model is trained and tested, it is stored in the model registry and then deployed in web service or IoT modules.
Image: It provides an environment to deploy the model independently. It consists of all the components required by the model. An image contains a model, application, or a script and dependencies required by the model or script. The images are stored in the image registry.
There are two types of images:
- FPGA image – used while deploying a field-programmable gate array in Azure ML. FPGA is a semiconductor device widely used in electronic circuits.
- Docker image – used to deploy computer targets such as Azure Kubernetes Service or Azure Container Instances. To know about docker images you can read our blog post on Docker Image: A Complete Guide For Beginners.
Deployment: The registered model is deployed as a service endpoint. It instantiates the image into a web service that is further hosted over the cloud or into an IoT module for using it in an integrated device deployment.
This is the entire workflow of an Azure Machine Learning Service.
Azure Data Scientist Associate Certification [DP-100]
If you are a data science enthusiast or studying machine learning then you should plan on taking the Microsoft Azure Data Scientist Associate [DP-100] Certification and must read our blog posts on [DP-100] Microsoft Certified Azure Data Scientist Associate: Everything you must know and know about the hands-on required by going through Microsoft Certified Azure Data Scientist Associate | DP 100 | Step By Step Activity Guides (Hands-On Labs).
Join FREE Masterclass
To know more about the AI, ML, Data Science for beginners, why you should learn, Job opportunities, and what to study Including Hands-On labs you must perform to clear [DP-100] Azure Data Scientist Associate.
Click on the below image to Register for our FREE Masterclass on Microsoft Azure Data Scientist Certification [DP-100] & Live Demo With Q/A Now!
The post Azure Machine Learning Service Workflow For Beginners appeared first on Cloud Training Program.